Wu Chuanqing

A robust anomaly based change detection method for time-series remote sensing images

Time-series remote sensing images record changes happening on the earth surface, which include not only abnormal changes like human activities and emergencies (e.g. fire, drought, insect pest etc.), but also changes caused by vegetation phenology and climate changes. Yet, challenges occur in analyzing global environment changes and even the internal forces. This paper proposes a robust Anomaly Based Change Detection method (ABCD) for time-series images analysis by detecting abnormal points in data sets, which do not need to follow a normal distribution. With ABCD we can detect when and where changes occur, which is the prerequisite condition of global change studies. ABCD was tested initially with 10-day SPOT VGT NDVI (Normalized Difference Vegetation Index) times series tracking land cover type changes, seasonality and noise, then validated to real data in a large area in Jiangxi, south of China. Initial results show that ABCD can precisely detect spatial and temporal changes from long time series images rapidly.

Optical properties and retrieving of diffuse attenuation coefficient of the Three Gorges Reservoir by remote sensing

By using the in situ measured data of spectral reflectance and water quality parameters in Aug.22-24,2009,this study analyzed the spectral characteristics and affecting factors of diffuse attenuation coefficient(Kd) of the Three Gorges Reservoir,and built a retrieving model of Kd(490)at band 490nm.In addition,Kd(490)was inversed by using CCD image of HJ-1 satellite in July 21,2009.The results showed that the attenuation peaks were not distinct at bands 440nm and 675nm owing to the high content of non-pigment particulates in the water.The dominant affecting factors were absorption of particulates and total scattering of water body,while average contributions of the four bands(i.e.440,550,613 and 675nm) were 45.48% and 38.21%,respectively;Kd strongly power correlated with the concentration of non-pigment particles between 400 to 700nm.Besides,the average relative error of inversed vs.measured Kd(490) was 17.89% for ground-measured Rrs,whereas the relative error for images data was 28.38%;Kd(490)of the study area was concentrated from 6 to 7m-1,which accounted for 83.60% of the study area.Additionally,Kd(490)of the tributaries of the reservoir was lower than the main stream,which indicated that the content of particulates in tributaries was relatively low.